Ok, I'm giving up trying to track down that floating point exception bug,
and reverted CVS back to the original code, with the following fixes:
VBR_min/max bitrate index set properly (Robert)
s3_s spreading function bug fixed (Alfred Weyers)
npart_l/npart_l_orig problems fixed? This was causing one of the
floating point exceptions.
Takehiro's code, which is faster and improves the logic of the
noise & distortion calculations can be checked out with:
cvs update -r takehiro2 quantize.c quantize-pvt.h loopold.c
("cvs update -A" to go back to main CVS version)
After that scalefac indexing bug was fixed, this code now passes all
my tests (when compiled w/o -DABORTFP), but I still get floating point
exception errors when compiled with -DABORTFP. As soon as this
problem is tracked down, these will be merged back into the main
branch. So if possible, hold off on committing changes to these three
routines. If you really want to commit some changes, track down the
exception for me!
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
